Study | Sample | Predictors variables | Outcome variables | Best performing ML approach | Explanatory power of best performing model | Limitations |
---|---|---|---|---|---|---|
Buscema et al. [49] | 172 females with a diagnosed ED | 124 different variables: generic information, alimentary behaviour, eventual treatment and hospitalization, substance use, menstrual cycles, weight and height, hematochemical and instrumental examinations, psychodiagnostic tests | ED diagnosis | Feed forward neural networks | 87% | Cross-sectional study Small sample |
Forrest et al. [27] | 191 adults with BED in a randomised controlled trial | Treatment condition, demographic information, baseline clinical characteristics | (1) Binge eating abstinence; or (2) reduction; (3) ED psychopathology; (4) perceived weight loss; and (5) actual weight loss | Elastic net | (1) 51%; (2) 4%; (3) 27%; (4) 12%; (5) 68% | Cross-sectional study Small sample |
Haynos et al. [26] | 415 female adults with a diagnosed ED, of which 320 completed measures at Year 1, and 277 completed measures at Year 2 | Demographics, psychiatric treatment, ED symptoms, other psychiatric diagnoses and symptoms, self-esteem | (1) ED diagnosis; (2) objective binge eating; (3) compensatory behaviours; and (4) underweight BMI | Elastic net regularized logistic regressions | At year 1: (1) 62%; (2) 77%; (3) 88%; (4) 93% At year 2: (1) 61%; (2) 71%; (3) 85%; (4) 89% | Small sample |
Krug et al. [33] | 1402 adolescents and adults (92% female), with (n = 588) or without (n = 760) a diagnosed ED | Cross-cultural risk factors for EDs before the age of 12 | (1) ED onset; (2) differential ED diagnoses | Penalised logistic regression (LASSO) | (1) 89%; (2) 70% | Cross-sectional study |
Linardon et al. [30] | 1341 adults (91% females), with (n = 512) and without (n = 829) recurrent binge eating | Intuitive eating behaviours, flexible restraint behaviours, rigid restraint behaviours, rigid restraint cognitions | Recurrent binge eating behaviour | Decision tree classification | 70% | Cross-sectional study |
Orru et al. [31] | 107 females, with (n = 53) and without (n = 54) a diagnosed ED | Presence of manic/hypomanic and depressive symptoms, AN and BN symptoms | ED status | Naïve bayes | 91% | Cross-sectional study Small sample |
Ren et al. [34] | 830 non-clinical young females | Psychological distress, eating inflexibility, body image inflexibility, body dissatisfaction, emotional overeating, loss of control overeating, body mass index | ED risk | Decision tree classification | 85% | Cross-sectional study |
Rosenfield and Linstead [32] | 44 female young adults, with (n = 20) and without (n = 24) a previous diagnosis of AN | ED symptoms, psychosocial impairment, symptoms of autism spectrum disorder | ED status | K-means clustering | 78% | Cross-sectional study Small sample |